Separation of Static and Non-Static Features from Three Dimensional Datasets: Supporting Positional Location in GPS Challenged Environments

摘要

Environments where Global Positioning Satellite (GPS) signals are intermittent or non-existent exist in a number of situations. Urban canyons, underground facilities, and building interiors are all well-known challenges to navigation due to the lack of consistent GPS signal lock. A workaround to this concern is critical for a wide ranging set of applications: navigation of autonomous vehicles, tracking of personnel in hazardous areas, and urban traffic routing for automobiles are all pertinent examples. Recent developments in Laser Radar (LADAR) technologies present a new opportunity to provide positional information in GPS challenged conditions. Emerging Flash LADAR technology provides the ability to acquire near continuous three-dimensional datasets coupled with positional information by existing technologies such as GPS or Inertial Navigation Units (INU). By comparing features in an image with known coordinates with the same features in images acquired from unknown positions, we can obtain an ongoing positional reference to establish location. A critical component of this methodology is the identification of features exhibiting a static nature and the exclusion of non-static features from the locating process. In this paper, research on the utilization of 3D feature matching to separate static and non-static features from a time series of LADAR images is presented. In particular, our findings on the use of polynomial signatures to match features between subsequent images are discussed. Finally, the potential for establishing camera position along with the associated uncertainties in space is demonstrated.